viktor/infra
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

tripit: mail-ingest extracts with the qwen3-8b text model, not the vision model

Browse source 
Forwarded schedule-change emails were being parsed by qwen3vl-4b (a 4B *vision*
model) for text extraction, which reliably dropped the flight number — so the
matcher had no key to link on and a forwarded flight update created a duplicate
instead of amending the existing segment.

Point the ingest-plans CronJob's text extraction at qwen3-8b (verified live: it
emits flight_number + a clean PNR, 3/3 on the failing email) and keep qwen3vl-4b
for boarding-pass image attachments (LLM_VISION_MODEL). llama-swap loads each on
demand; the GPU swap cost is accepted.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
Viktor Barzin 2026-06-16 20:39:29 +00:00
parent 044444d328
commit 14476bfbd7
1 changed files with 8 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

11
stacks/tripit/main.tf
View file

@ -550,9 +550,14 @@ locals {
suspend = false
imap_pw_secret_key = "PLANS_IMAP_PASSWORD"
extra_env = {
LLM_MODE = "llamacpp"
LLM_ENDPOINT = "http://llama-swap.llama-cpp.svc.cluster.local:8080"
LLM_MODEL = "qwen3vl-4b"
LLM_MODE = "llamacpp"
LLM_ENDPOINT = "http://llama-swap.llama-cpp.svc.cluster.local:8080"
# Text body extraction uses the 8B text model (reliably emits flight_number);
# boarding-pass image attachments use the 4B vision model. llama-swap loads
# each on demand. Was qwen3vl-4b for both, which dropped flight numbers and
# duplicated schedule-change emails (2026-06-16).
LLM_MODEL = "qwen3-8b"
LLM_VISION_MODEL = "qwen3vl-4b"
MAIL_INGEST_ENABLED = "true"
IMAP_HOST = "mailserver.mailserver.svc.cluster.local"
IMAP_PORT = "993"